It Takes Two to Tango: Process Integration and Wages

Transcription

1 (preiminary; comment ecome) It Take To to Tango: Proce Integration and Wage by Aberto Damazzo* and Paquae Scaramozzino** June 2000 * Dipartimento di Economia Poitica, Univerità degi Studi di Siena, Piazza S. Franceco 7, I Siena. damazzo@unii.it ** Centre for Financia and Management Studie, SOAS, Univerity of London, Tornaug Street, London WC1H 0XG. p6@oa.ac.uk Abtract Ti paper ook at te reationip beteen tecnoogica compexity and age. Uing a pane of UK etabiment, e find tat age increae it te degree of compexity, meaured bot by te tak ratio and by te preence of integrated proce contro. Our finding are conitent it Kremer' (1993) "O-Ring" production function, ic impie trong compementaritie among te production activitie. Te abiity to impement performance appraia ytem a a negative and ignificant effect on age, conitent it te efficiency age ypotei. Te ue of minicomputer or te exitence of a computer netork increae te itin-firm age diperion, ic i ao conitent it te preence of compementaritie in te production proce. Keyord: tecnoogica compexity, proce integration, age diperion JEL Caification: J31, O33

2 1. Introduction Te at decade ave itneed important cange bot in te production tructure and in age diperion, eading many oberver to concude tat ne tecnoogie are coey reated to riing age inequaity. Te mot prominent ypotei i tat te introduction of advanced tecnoogie a caued a trong increae in te demand for kied orker 1. Furter, Kremer and Makin (1996) ave uggeted tat tecnoogica cange may ave ubtantiay contributed to increaing egregation among orker of different quaity into ig-age and o-age firm. Le attention a been paid to te reation beteen tecnoogica cange and increaing itingroup age inequaity, or reidua variation, moty attributed to unoberved orker abiity 2. Hoever, according to ome recent reut by Abod, Kramarz, and Margoi (1999), individua abiity aone cannot entirey expain age differentia: empoyer effect are found to be ti reevant. Te cope of ti paper i to provide a teoretica and empirica anayi, at te firmeve 3, of te reation beteen compex tecnoogie and age-tructure. In particuar, our approac can expain y identica orker may receive different age in equiibrium, and y compex tecnoogie may increae age inequaity among orker of different quaity. Te firt part of te paper deveop a teoretica mode ere tecnoogy i repreented by te Kremer' (1993) O-Ring production function. Production require te execution of a certain number of riky tak. Te number of tak i a meaure of tecnoogica compexity ince, if even a inge tak i mi-performed, output i ot. Ti notion of tecnoogy impie te exitence of trong compementaritie among te tak, or activitie, tat contitute te production proce. Simiar idea are put forard ao by Migrom and Robert (1990, 1995), o argue tat te introduction of advanced tecnoogie uc a Computer-Aided Deign or Computer-Aided Manufacturing (CAD/CAM) ytem a increaed te degree of interdependence among te productive activitie of te firm. We o tat, en monitoring i imperfect, te empoyer i ave an incentive to pay age tat induce empoyee to eicit a proper eve of effort. Te adoption of efficiency age in ti context i pecuiar to our mode. We ao o tat, en te degree of compexity in production i reativey ig, te empoyer i pay iger age on average. Furtermore, by conidering production procee ere different tak ave a different enitivity to effort, e 1 See, among many oter, Berman et a. (1994), Autor et a. (1998), Dom et a. (1997), Godin and Katz (1998), and Hake and Heden (1999). 2 On itin-group inequaity, ee for exampe Gottcak (1997) and Acemogu (1999). 3 Te abour-market impication of te baic teoretica mode expoited ere are expored in Damazzo (2000). 1

3 o tat orker o poe te ame abiity can be paid different age depending of te tak tey are aigned to. Tu, te preent mode i particuary uefu to expain bot beteen-firm and itin-firm age inequaity among identica orker 4. Te econd part of te paper (Section 3) preent ome cro-ectiona evidence on te reation beteen tecnoogy and age dran from WIRS 1990, a urvey of Briti etabiment. Te aim of our empirica anayi i tofod. Firt, e eek to ae te impact on age of proxie for tecnoogica compexity, uc a ue of microproceor tecnoogy, R&D expenditure, and te proportion of compicated tak. Second, e try to evauate te impact on te age tructure of meaure of inter-dependence in production activitie, ince te O-Ring mode i caracteried by compementary tak. Te reut e obtain are conitent it te prediction of te teoretica mode. On te one and, e o tat firm tat adopt reativey compex tecnoogie pay ig age to eac category of orker conidered (unkied, emi-kied, kied, cerica, upervior). On te oter and, e o tat differenty from mot of te iterature on computer and age not a form of microproceor uage are reated to iger age. Hoever, conitenty it te notion of te O-Ring tecnoogy, e find tat te ue of integrated proce contro ytem i poitivey and ignificanty reated to age. Our anayi a ome obviou imit, ince e ave no meaure for te orkforce abiity. For ti reaon, e cannot contro for orting of ig-quaity orker into tecnoogicay-advanced firm, a te competitive mode deveoped by Kremer and Makin (1996) ugget. Hoever, our empirica reut are conitent it te efficiency-age ypotei tat form te bai for our prediction. In fact, e find tat te abiity to impement performance appraia ytem itin te firm ic e interpret a a proxy for te abiity to monitor empoyee a a negative and ignificant effect on age. Te paper i compoed a foo. Section 2 deveop te baic teoretica mode, and extend it to te endogenou coice of monitoring. Section 3 preent te empirica reut on te reation beteen tecnoogy, age eve, and itin-firm age diperion. Section 4 concude. 2. Te mode In te fooing ection, e deveop te baic mode of te firm optima beaviour en te production function a te O-Ring form and orker can be imperfecty monitored. In ection 2.2, e extend te baic mode to endogenie te coice of monitoring. In ection 2.3, e 4 On itin-firm inequaity, ee Entorf et a. (1999) and te reference quoted terein. 2

4 examine te effect of product market competition by caracteriing a zero-profit equiibrium, and obtain te main prediction of te teory Te baic frameork. In te mode e deveop, orker diike effort and monitoring i imperfect. A in te efficiency-age mode by Summer (1988), e aume tat te eve of effort e exerted by a orker i given by te fooing function: x e = x 0 β if if > x x (1) ere i te age paid, x denote te vaue of orker abour-market aternative, and β (0,1). For at concern te firm tecnoogy, e borro from Kremer (1993) te notion of O-Ring production function. Production require tat n tak be correcty performed. Here, te probabiity tat a tak i carried out uccefuy depend on te effort put in by te orker o i in carge of it. Simiary to Kremer and Makin (1996), production invove a number of eay tak ( n), and (n-) ard tak. We aume tat eay and ard tak ave a different enitivity to effort. In particuar, an eay tak i be performed correcty it probabiity equa to q=q(e), it q' >0. On te oter and, a ard tak i be performed correcty it probabiity q z =[q(e)] z, ere z>1. Tu, given te eve of effort put in, a ard tak i e ikey to ucceed tan an eay tak. Witout o of generaity, e aume tat eac tak i performed by a inge orker. Wen: (1) tak ave different difficuty, (2) a te orker aigned to ard tak put in te ame eve of effort, and (3) a te orker aigned to eay tak put in te ame eve of effort, ten te O-Ring production function can be ritten a foo: y = n B( n, ; f ) K α q q z( n) (2) ere K i te firm capita eve, and α B( n, ; f ) K denote te average per-orker revenue en a te tak are correcty performed, an event occurring it probabiity q q z( n). Te 3

5 variabe f denote te number of identica firm in te indutry, and e take B f <0: te iger te number of competing firm, te oer te firm revenue. Note tat te production function (2) entai compementarity among tak: a i be on, te optima age to be paid for a certain tak depend on te age eve paid for a te oter tak, be tem eay or ard. Ti property foo directy from te tecnoogy conidered: in te mot extreme interpretation of Kremer (1993), te degree of integration among tak i uc tat, if ony one tak i mi-performed, te vaue of production drop to zero. Atoug not neceary for our main argument, it can be uefu to tink of te mode a if tere ere to type of orker: ig-kied (indexed by ) and o-kied (indexed by ), auming tat orker of different ki cannot be ubtituted for one anoter 5. Ten, o-ki orker i be aigned to eay tak, ie ig-ki orker i be aigned to ard tak. For exampe, manageria tak i be given ony to orker o meet particuar requirement in term of education, experience, etc. Wen tere are common preference toard abour diutiity, equation (1) i od for bot type of orker, it x and x, denoting repectivey te ig-kied and te o-kied abour-market opportunitie. In cae tat a orker poe te ame ki, it i od tat x =x. For impicity, e take te probabiity q to be equa to e. Hence, te firm maximum probem take te form: max Π {,, K} = nb( n, ; f) K α x x β x x zβ ( n) ( n) rk F (3) ere F 0 repreent a fixed cot in production. Te firm maximie profit by cooing imutaneouy te age eve and paid to te orker aigned to eay and ard tak repectivey, and te eve of capita K. Te oution to probem (3) generate te fooing et of firt-order condition: Π = β y x 1 = 0 (4) Π = z β y x 1 = 0 (5) 4

6 Π = K α y r K = 0 (6) ere y i te expected eve of production, a defined by (2). Note tat te equiibrium age eve depend, troug y, on te equiibrium age eve, and vicevera. By manipuating te condition (4) and (5), one obtain tat: x x = z (7) Condition (7) impie tat ( - )>(x -x ). Tu, if te kied orker outide opportunitie are not ore tan te unkied orker one (ic i, if x >x ), te firm i pay kied orker a iger age. Te mode ao capture te particuar cae en te firm ire orker of te ame quaity to perform tak of different compexity. Ten, it od tat x =x, and condition (7) impie tat orker aigned to arder tak i be paid more. Tu, even itin te ame firm, orker of equa ki can be paid different age. Ti carifie tat at drive age diperion ere i not orker quaity, but te fact tat different tak itin a firm are differentiay enitive to effort. We no concentrate on te effect of n and on te equiibrium vaue of age and capita, tat i, on,, and K. Te number of tak n can be interpreted a a meaure of compexity in production: given, te greater te number of tak, te greater te probabiity tat ometing goe rong and production i ot. On te oter and, given n, a greater number of eay tak reduce te eigt of ard, and rikier, tak on te production proce. By differentiating te ytem (4)-(6), and expoiting te econd-order condition aociated it probem (3), one obtain te fooing: Reut 1. An increae in te number of tak n i raie te age paid to ig-kied orker enever expected revenue are increaing in n: ence, it od tat d /dn>0 enever dy/dn>0. Simiary, it od tat d /d>0 enever dy/d>0. 5 By contrat, Kremer and Makin (1996) conider orker o ave intrinicay different productivity and anayze te optima matcing deciion of te firm. 5

7 Note ao tat en d /dn i poitive, bot d /dn and dk/dn are poitive a e, ince ( -x )=z ( -x ) and K = α ( x ) /( zrβ ). Anaogouy, en d /d>0, it i od tat d /d>0 and dk/d>0. Tee impication of Reut 1 need ome dicuion, becaue tey cruciay depend on te ign of dy/dn and dy/d. Wat are te pauibe ign for dy/dn and dy/d? Conider firt te cae of a firm ic i conidering eter to impement a more compex tecnoogy (iger n). Te tecnoogy i be adopted ony en profitabe, tat i, ony en te condition dα/dn>0 od. By te enveope teorem, it od tat dα/dn=dy/dn- *. Conequenty, it mut be true tat dy/dn>0. We can concude tat increaingy compex tecnoogie i ony be impemented en revenue y are increaing in n: ti can occur for exampe en more opiticated product e at iger price. Tu, a impied by Reut 1, te adoption of more compex tecnoogie i bound to raie age. We can no dicu te effect of a cange in. Given te degree of compexity in te tecnoogy conidered (i.e., given n), a rie in impie tat ome ard tak can be tranformed into eay, e riky tak. Ti occur, for exampe, en certain tak requiring a ig eve of ki can be reduced into routine performed by e paid orker 6. On baance, te net effect of cange in on y i intrinicay ambiguou. On te one and, a greater number of eay tak, given n, reduce te rik and increae revenue expected vaue. On te oter and, it i reaonabe to expect tat te condition B <0 od: product tat rey on very tandardized production metod are ikey to e at oer price. A on in Section 2.3, more precie prediction on te effect of n and on ageeve can be obtained by conidering competition among firm Coice of monitoring. Te mode deveoped in te preceding ection can be extended to te cae en te firm dicipine it orker not ony by etting te age eve, but ao by cooing te eve of monitoring. Higer monitoring intenity faciitate te detection of orker i-performance. To conideration make monitoring reevant. Firt, monitoring and, more in genera, uperviion a a reevant roe in te organization of production (ee, among oter, Buo and Summer (1986)). Second, tere remain te teoretica poibiity tat, en te degree of tecnoogica compexity increae, firm react by increaing teir monitoring intenity intead of 6 See te exampe of te printing indutry reported in Mark (1987). See ao Brenaan (1999) on ite-coar tak. 6

8 7 raiing age. We aume tat te firm can cooe to monitor a orker aigned to a certain tak it intenity m. Te cot of m i given by M(m), it M >0 and M 0. Since iger monitoring intenity i induce eac orker to exert more effort, e can modify te effort function (1) a foo: > = oterie 0 0 ), ( if ) (1 m x m x x e β σ σ (8) ere =(,) denote te orker type and σ (0,1). Te maximum probem no take te fooing form: { } ( ) ( ) F K r n m M m M y K m m + + = Π ) ( ) ( ) ( ˆ max,,,, (9) ere ) ( ) (1 ) (1 ) ;, ( ˆ n z m x x m x x K f n B n y = β σ σ β σ σ α By cacuating and re-arranging te firt-order condition of probem (9), one obtain: ), (, ) (1 ) '( m M m x = = σ σ (10) Note tat, if te degree of convexity of te function M(m) i trong, te firm i rey mainy on te age to induce orker to perform adequatey. Wen te cot of monitoring i inear, tat i, en µ = m m M ) (, condition (10) impie tat any cange tat affect te age-premium ( - x ) i vary te coice of monitoring intenity m in te ame proportion: σ µ σ = 1 m x (11)

9 Te iger te margina cot of monitoring, µ, te iger i te age reative to monitoring intenity. Ao, en te abiity to monitor i rater imited, te empoyer i pay iger age. In concuion, monitoring doe not generay eiminate te need to pay ig age en compex tecnoogie are adopted. For ti reaon, e i abtract from monitoring in te ret of te teoretica mode Te effect of competition. Reut 1 od for a given number of firm, f. In ti ection e conider product-market competition troug te entry of ne firm, in order to obtain arper prediction from firm-eve comparative tatic. To ti purpoe, e impoe tat te zero-profit condition od. Ti condition take te form 7 : y = + ( n ) + r K F (12) + Expoiting condition (12) togeter it equation (4)-(6), one obtain te fooing ytem of tree equation in (,,K): β [ + ( n) + rk + F] ( x ) = 0 (13) zβ [ + ( n) + rk + F] ( x ) = 0 (14) α [ + ( n) + rk + F] rk = 0 (15) By totay differentiating te ytem (13)-(15) it repect to (,, K, n,, F), e obtain te fooing reut 8 : Reut 2. Under te zero-profit equiibrium, it od tat: (i) d /dn, d /dn, and dk/dn are poitive, (ii) d /d, d /d, and dk/d are negative, and (iii) d /df, d /df, and dk/df are poitive. 7 Te zero-profit condition determine f*, te equiibrium number of firm operating in te market. 8 Te determinant aociated to te ytem (13)-(15) i negative. Te proof of Reut 2 i rater immediate, atoug it require tediou cacuation. 8

10 Reut 2 a ome reevant impication. Firt, en a firm adopt a reativey compex tecnoogy (i.e., en n i ig in our mode), it i be ready to pay reativey ig age to it empoyee, be tem attaced to ard or eay tak. Higer compexity raie te rik of faiure in production. Tu, ince tak are compementary, te empoyer i ave an incentive to eicit more effort from every empoyee. In ti perpective, age inequaity tend to arie mainy troug pant egregation : tere are ig-age pant and o-age pant according to te type of tecnoogy adopted. 9 Many contribution on age ave identified increaing opitication in tecnoogy it te diffuion of computer, and microproceor tecnoogy in genera, on te orkpace: ee Krueger (1993), Berman, Bound, and Griice (1994), Berman, Bound, and Macin (1998), Dunne, Foter, Hatianger, and Troke (2000), among many oter. In particuar, Autor, Katz and Krueger (1998, p.1168), and Brenaan (1999) ave argued tat computer eem to be better uited to eiminate eay tak, rater tan compex and idioyncratic tak 10. In teir cro-ection anaye on age, Dunne and Scmitz (1995) and Dom, Dunne and Troke (1997) ave ued te number of ne factory automation tecnoogie to proxy for production proce compexity. Van Reenen (1996), and Macin and Van Reenen (1998), among oter 11, ave eected R&D expenditure a a meaure of tecnoogica opitication. Even from our perpective, tee meaure remain uefu to proxy for compexity. Hoever, conitenty it te O-Ring approac, e i ao need ome tecnoogica variabe to account for te degree of inter-dependence among tak. Reut 2 a a econd important impication for te firm age poicy, ic i reated to te proportion beteen ard and eay tak. Firm tat experiment a decreae in te proportion of eaier tak, a meaured by a reduction in for given eve of n, i pay teir orker more, be tem attaced to ard or eay tak. An increae in ratio beteen ard tak and eay tak i at Jonon (1997, p.48) a termed extenive tecnoogica cange : our mode predict tat ti type of tecnica cange i ead te empoyer to pay iger age to a te empoyee. Intuitivey, an increae in te proportion of ard tak ead to iger rik of production faiure for any given eve of effort put in by orker. Conequenty, te empoyer a an incentive to pay iger age to eicit more effort. Reut 2 a ao a tird impication. Firm tat bear ig fixed cot, a meaured by F, tend to afeguard teir invetment by paying ig age, o a to obtain a better performance 9 For evidence on pant-egregation, ee Davi and Hatianger (1991) and, in particuar, Kremer and Makin (1996). 10 On ti iue, ee ao Jonon (1997). 9

11 from teir orker. Fixed cot can capture different kind of unk invetment, uc a irreveribe equipment, or R&D itef. Moreover, te amount of fixed cot i ikey to be reated to te firm ize, meaured by number of time te baic production proce i repicated. Te idea tat empoyer caracteritic, uc a te feature of tecnoogy, drive te firm age-poicy a been idey dicued. A argued by Krueger and Summer (1987,1988), tecnoogica factor eem to be very reevant to expain age tructure over different countrie. Moreover, Katz (1986) and Gibbon and Katz (1992) ave noted tat inter-indutry age reativitie are remarkaby imiar in different occupation. On te bai of Katz evidence, Layard, Nicke, and Jackman (1991, p.167) ave concuded tat ti i not at a conitent it a tecnoogica approac. Wy oud te reponibiitie of office orker, janitor, tecnician, and operative in different indutrie a vary in proportion? It eem mot unikey. 12 Hoever, te objection raied by Layard et a. (1991) i not vaid en tecnoogy i repreented by an O- Ring production function. In te O-Ring perpective, production ucceed en a number of inter-dependent operation are performed correcty. Conequenty, i-performance by an eectrician in te production of airpane can ave more dramatic conequence tat i oppy performance in te car indutry. Te ue of te O-Ring production function i conitent it aternative age teorie. According to te efficiency-age approac e foo, compicated tecnoogie induce firm to pay ig age, o a to improve performance. By contrat, te competitive approac fooed by Kremer and Makin (1996) predict tat firm adopting compex tecnoogie i ire igki, ig-age orker. Hence, increaingy compex tecnoogie may ead to kiegregation among pant 13. Note oever tat, if ki ere te ony reevant factor and monitoring probem ere irreevant, te preence of incentive ceme oud be difficut to expain. Anoter impication of te mode i reated to te age gap beteen orker performing ard tak and orker performing eay tak. Te fooing reut od: See ao Berman et a. (1994). 12 A imiar point i made in Gibbon and Katz (1992). 13 Te mode in Kremer and Makin (1996) i conitent it expanation baed on ki-biaed tecnica cange and, more in genera, it te idea of ki-tecnoogy compementarity: ee Godin and Katz (1998). 14 To verify Reut 3, note tat d( )/dn=z (d /dn). 10

12 Reut 3. and. An increae in n, or a decreae in, i raie te age differentia beteen Ti Reut a intereting impication for te age poicy fooed itin a firm, and in particuar for at concern itin-firm inequaity. Increaing itin-firm age inequaity a been oberved by Davi and Hatianger (1991) in US manufacturing pant, and by Kramarz, Loivier and Peè (1996) for France. According to Entorf, Goac, and Kramarz (1999), ti penomenon i accounted for at eat in part by te introduction of ne tecnoogie. To ummarize, te impe mode preented may account for to type of age inequaity: (i) Witin-Group age inequaity. Our mode can expain y identica orker can be paid different age in equiibrium. Suppoe tat a orker poe omogeneou ki. According to Reut 2, firm tat adopt compex tecnoogie i pay teir orker iger age tan firm tat ue impe tecnoogie (itin-group, beteenpant inequaity). Furtermore, Reut 3 predict tat identica orker can be paid different age by te ame empoyer depending on te tak tey are aigned to (itingroup, itin-firm inequaity). Te tyied aumption tat a orker poe te ame ki i patenty extreme, but it neaty o te pecuiar impication of our mode it repect to oter. Our mode can fuy account for itin-group age diperion, ince it depend on empoyer effect reated to tecnoogy. By contrat, competitive mode treat reidua variation merey a unoberved abiity. At te ame time, it i quite reaonabe to uppoe tat ard tak are ikey to be given to ig-quaity orker. Under ti preumption, our reut can expain age inequaity among eterogeneou orker: (ii) Beteen-Group age inequaity. Firm tat adopt rikier tecnoogie a meaured by a iger n, or a oer i exibit greater age diperion (ee Reut 3). Conequenty, te more compex te tecnoogy, te greater te eve of age inequaity among ig-ki and o-ki empoyee. 11

13 3. Empirica anayi Te teoretica impication of te mode are teted in at foo. Section 3.1 preent te empirica trategy. Section 3.2 decribe te data e ue. Finay, Section 3.3 preent and dicue te empirica finding An empirica mode Te teoretica mode et out in ection 2 a important empirica impication. Bot te eve and te ditribution of age in te firm are reated to te compexity of te production proce. Reut 2 impie tat te age of a categorie of orker oud be poitivey affected by te reative proportion of compex tak in production. Reut 3 predict tat te variabiity of age at te firm eve oud be an increaing function of te reative proportion of compex tak. Let x denote a vector of contro for age and u j, j=1, 2, 3, denote a tocatic diturbance. We can rite te fooing empirica reationip: = f 1 ( n /, x, u 1 ) (16) 2 2 = f ( n /, x, u ) =, (17) σ 2 = f 3 ( n /, x, u 3 ) (18) Equation (16) ay tat te average age in te firm ( ) i a tocatic function of te ratio of compex tak, n/, and of a vector of contro variabe, x. According to Reut 2, te average age i an increaing function of te tak ratio: / ( n / ) > 0. Equation (17) ay tat, ceteri paribu, te age of every category of orker i an increaing function of te tak ratio: / ( n / ) > 0, =,. Bot equation (16) and (17) are impied by Reut 2. Equation (18) ay tat, after controing for te oter factor affecting age, te variance of age at te firm 2 eve, σ, i an increaing function of te compexity of production a proxied by te tak ratio: 2 σ / ( n / ) > 0. Equation (18) i impied by Reut 3, ic ay tat te age differentia 12

14 beteen ig- and o-compexity job i an increaing function of te ratio of ig- to ocompexity tak in te firm. Apart from te variabe tat capture te roe of compexity or proce integration, equation (16)-(18) oud incude te fooing main categorie of regreor: tecnoogica variabe and abour/union variabe. Te tecnoogica variabe can capture te productivity difference acro firm tat are not reated to te tak ratio. Te abour and union variabe coud capture bot te firm' abiity to monitor it empoyee and te reative bargaining poer of orker in age etting. In addition, e can incude indutry dummie to contro for a te oter difference acro indutrie for ic e do not ave expicit information Te data Te data et conit of a ampe of 284 UK etabiment from te 1990 Workpace Indutria Reation Survey (WIRS). Ti data et i particuary attractive ince it contain information on a number of tecnoogica and union variabe at te etabiment eve. In addition, WIRS provide information on financia variabe and on product market caracteritic. We imit our ampe to etabiment operating in te trading ector: governmenta organiation and nontrading pubic corporation are terefore excuded from our anayi. Te data e ue in our anayi are ummaried in Tabe 1. Te dependent variabe are te median age of unkied, emi-kied, kied, cerica and upervior empoyee. We ao contruct te average oury age a te eigted average of te above age by category of orker. Our meaure of age diperion i te um of te proportion of empoyee earning e tan af te average age or more tan tice te average age in te etabiment. Te main variabe meauring te compexity in production i te tak ratio, defined a te proportion of midde/enior manager and enior tecnica/profeiona on tota empoyment. Ti can be een a capturing te proportion of ard tak over tota tak. Te teory predict a poitive effect on a age, incuding te age paid by te firm to te orker o perform eay tak. Anoter critica variabe i te preence of integrated proce contro (IPC). Ti i defined a te microeectronic contro of integrated procee, ere te contro invove evera tage of production procee. Ti variabe meaure te intenity of te interdependency acro tak. A iger degree of integration exacerbate te conequence on tota production of mitake in a inge tak. Te effect on age oud be poitive for a categorie of orker. 13

15 We ao incude to variabe reated to computer uage: te preence of minicomputer and te exitence of a computer netork. Te firt variabe can be een a mainy capturing te compexity of tak in te etabiment, and te econd variabe a mainy capturing te degree of interdependency acro tak. Bot variabe meaure te ue of advanced tecnoogy and compementarity among production tak. Te proportion of tota expenditure of te organiation pent on R&D can ao be een a a meaure of te compexity of tak in production. Among te tecnoogica variabe, e incude meaure of reative productivity. Tee meaure conit of a comparion of abour productivity in te etabiment it oter imiar orkpace. Anoter important tecnoogica variabe i te ize of te etabiment, ic e meaure by tota empoyment. Te ize of te etabiment i ao important for: (1) te abiity of te firm to monitor it empoyee: te bigger te ize, te more difficut it i to monitor orker (te expected ign on age i terefore poitive); (2) a meaure of fixed cot: arger firm tend to be aociated it arger fixed cot (te mode again predict a poitive effect). Among te abour variabe, union effect are meaured by te proportion of unionied empoyee on tota empoyment. Ti variabe can infuence te outcome of bargaining over te age above te efficiency age eve (ic oud ten pace a oer boundary on te range of age). We expect a poitive ign. We ao conider te exitence of performance appraia ytem, i.e. individuay ritten aement produced periodicay by manager or upervior. Appraia ytem are ikey to be aociated it iger monitoring abiity. It i terefore poibe to eicit a given eve of effort it a oer eve of age. Te ign predicted by te teory i terefore negative. An additiona abour contro i turnover: a iger turnover coud ead to iger efficiency age being paid. We incude one-digit indutry dummie in our regreion to capture any inter-indutry ource of difference in age. Our regreion terefore capture te intra-indutry tructure of age eve. Additiona contro in our empirica regreion ere meaure of output demand eaticity, product caracteritic, foreign onerip, market tructure indicator, entry barrier, financia performance, joint conutative committee, capacity utiiation, and uage of pecia too and macinery. 14

16 3.3. Empirica finding Tabe 2 preent regreion of te average oury age. Coumn (1) incude etabiment ize (SIZE), te unioniation rate (UNION), te exitence of appraia ytem (APPRAISAL), te tak ratio (TASK RATIO), te preence of integrated proce contro (IPC), and te interaction beteen te tak ratio and integrated proce contro. A variabe are ignificant and ave te expected ign. Size, ic may capture bot te abiity to monitor and fixed cot, a a poitive effect on te average age, a expected. Te unioniation rate a a poitive effect on te average age. Te abiity to monitor meaured by te exitence of an appraia ytem reduce te average age by about 11%. Te mot important variabe for u i te preence of integrated proce contro. Teir margina effect i poitive and etimated at 0.09 at te ampe average. On average, tu, te preence of integrated proce contro increae te average age by about 9%. Coumn (2) ao incude expenditure on R&D, bot on it on and interacted it te tak ratio. Te margina contribution of R&D i poitive en etimated at ampe average, but not tatiticay ignificant. Te coefficient on te oter variabe are not affected by te incuion of R&D. Coumn (3) and (4) incude additiona contro: tee are eter te productivity in te etabiment i iger or oer en compared it imiar orkpace, and te turnover ratio. Te productivity variabe ave te correct ign, but tey are not tatiticay ignificant, bot itout and it R&D. A iger turnover i aociated it a oer average age, but again te regreor i not tatiticay ignificant. We ave ao incuded a number of aternative regreor, but tey are not tatiticay ignificant. In particuar, in contrat to oter empirica tudie (Krueger, 1993; Autor, Katz and Krueger, 1998; DiNardo and Picke, 1997) ic find tat computer ue a a poitive infuence on age, e find tat variou meaure of te adoption of computing faciitie ave no ignificant effect on te eve of age, once e contro for integrated procee. Dom, Dunne and Troke (1997), Dunne, Foter, Hatianger and Troke (2000) ue meaure of tecnoogica advancement at pant eve (uc a CAM/CAD) and find a poitive aociation it age. Among te aternative meaure of microeectronic appication tat e ue (uc a deign, contro of individua macine, contro of individua item of proceed pant, centraied macine contro of group of macine, IPC, automatic anding of product and materia, etc.), by far te mot important one i IPC. We ao find tat te ue of aternative type of too and macinery ave no ignificant effect on age. 15

17 We oud note tat e cannot contro for te abiity or ki of orker, but indutry dummie can party capture te orting of ki among production ector (Barte and Sicerman, 1999). Tabe 3 preent te etimate correponding to coumn (2) of Tabe 2 for te median age of individua categorie of orker: unkied, emi-kied, kied, cerica and upervior. Te tak ratio, integrated proce contro and teir interaction remain ignificant for a group of orker. An important reut i tat te etimated coefficient for te tak ratio tend to increae it te degree of orker peciaiation. Te unioniation rate and te appraia ytem are bot quantitativey more important and more tatiticay ignificant for manua orker (unkied, emi-kied and kied). A poibe expanation for te atter finding i tat te abiity to monitor troug te appraia ytem decreae it te reative compexity of te tak. R&D tend to be quantitativey more important for cerica and upervior, atoug it i not tatiticay ignificant. Tabe 4 preent regreion on te age diperion at etabiment eve (itin-firm age diperion). Te dependent variabe i te um of te proportion of empoyee oe age i greater tan tice te average age in te etabiment, or e tan af te average age. Te dependent variabe i terefore a meaure of te variabiity of te ditribution of age in te etabiment. Coumn (1) o tat ize a a negative infuence on age diperion. Combining ti reut it tabe 2 and 3, e can infer tat, on average, arger firm pay iger age and dipay oer itin-firm age variabiity. Unioniation a a negative infuence on age, conitent it at e oud expect on te roe of union. Fooing our teoretica dicuion in ection 2, te microeectronic variabe e conider are te ue of minicomputer, te exitence of a computer netork, and te tak ratio. Minicomputer ave a poitive effect, conitent it greater compexity. Te preence of a computer netork a a poitive effect, conitent it enanced interdependency acro tak. Te tak ratio a a poitive infuence, a e oud expect from te teory, atoug it effect i not tatiticay ignificant. Coumn (2) and (3) preent parimoniou repreentation of coumn (1). Te effect of te variabe are confirmed en e ony incude minicomputer or te exitence of a computer netork. To ummarie te reut in Tabe 4, atoug our data do not confirm a direct effect of microeectronic on te eve of age (unike, for intance, Krueger, 1993, and Entorf, Goac and Kramarz, 1999), e do find tat te preence of a computer netork or te ue of minicomputer tend to increae te itin-firm diperion of age. Bot a computer netork and a mini-computer can be een a aociated it a greater degree of inter-dependency among te 16

18 tak carried out by individua orker. Our finding are terefore conitent it te vie tat te itin-firm age differentia increae it te reative compexity of tak in te firm. 4. Concuding remark Ti paper ook at te reationip beteen tecnoogica compexity and age. Uing a pane of UK etabiment, e find tat age increae it te degree of compexity, meaured bot by te tak ratio and by te preence of integrated proce contro. Ti i true bot for te average age in te etabiment, and for te age of eac category of orker. Our finding are conitent it Kremer' (1993) "O-Ring" production function, ic aume trong compementaritie among te production activitie. We ao find tat te abiity to impement performance appraia ytem a a negative and ignificant effect on age, conitent it te efficiency age ypotei. Te ue of minicomputer or te exitence of a computer netork increae te itin-firm age diperion, conitent it te preence of compementaritie in te production proce. We do not ave information on individua orker' quaity at te firm eve, and cannot terefore rue out tat more compex tak are aociated it iger age becaue of teir compementarity it ki (Godin and Katz, 1996, 1998). Our finding oever o tat te tak ratio exert a poitive infuence on te age of a categorie of orker, and are terefore conitent it te exitence of trong compementaritie among tak, a uggeted by te "O- Ring" tecnoogy. 17

19 Lit of variabe Te expreion in bracket denote te WIRS (1990) code. Tota empoyment; number of empoyee by category. (C1). Median pay. (K12). Hour per eek. (K13). Paid overtime our. (K14). Proportion of fu-time empoyee earning af te average gro earning or e. (K15). Proportion of fu-time empoyee earning tice te average gro earning or more. (K16). Number of empoyee o reigned during te at teve mont. (P1). Preence of integrated proce contro. (A26_07). Mini-computer ue. (A27_03). Preence of a computer netork. (A27_05). Proportion of organiation' tota current expenditure pent on reearc and deveopment. (FB13). 18

20 Tabe 1. Decriptive tatitic Log average oury age Log average oury age, unkied Log average oury age, emi-kied Log average oury age, kied Log average oury age, cerica Log average oury age, upervior Wage diperion Number of empoyee Union recognition Performance appraia ytem mean tandard median 25% percentie 75% percentie deviation Tak ratio Integrated proce contro Minicomputer Computer netork Reearc and deveopment Hig productivity Lo productivity Turnover

21 Tabe 2. Proce integration and average age. Dependent variabe: og average oury age (1) (2) (3) (4) SIZE ** ( ) * ( ) ** ( ) ** ( ) UNION ** ( ) ** ( ) ** ( ) ** ( ) APPRAISAL ** ( ) ** ( ) ( ) ( ) TASK RATIO ** ( ) ** ( ) ** ( ) ** ( ) IPC ** ( ) ** ( ) ** ( ) ** ( ) TASK RATIO *IPC ** ( ) ** ( ) * ( ) ** ( ) R&D ( ) ( ) TASK RATIO *R&D ( ) ( ) HIGH PROD ( ) LOW PROD ( ) TURNOVER ( ) ( ) ( ) ( ) Indutry dummie Ye Ye Ye Ye n F ** 4.07** 3.30** 2.77** F ** 4.73** 4.62** 4.64** R e Note: Heterocedaticity-conitent.e. in bracket. *: tatiticay ignificant at 10% eve; **: tatiticay ignificant at 5% eve. F 1 : joint ignificance of expanatory variabe. F 2 : joint ignificance of indutry dummie. 20

22 Tabe 3. Proce integration and age by category of orker. Dependent variabe: og average oury age Unkied Semi-kied Skied Cerica Supervior SIZE ** ( ) ( ) ( ) ** ( ) * ( ) UNION ** ( ) ** ( ) ** ( ) ( ) ( ) APPRAISAL ** ( ) ** ( ) * ( ) ( ) ( ) TASK RATIO ** ( ) ** ( ) ** ( ) ** ( ) ** ( ) IPC ** ( ) ** ( ) ** ( ) ** ( ) ** ( ) TASK RATIO *IPC * ( ) ** ( ) ** ( ) ** ( ) ** (0.4698) R&D ( ) ( ) ( ) ( ) ( ) TASK RATIO *R&D ( ) ( ) ( ) ( ) ( ) Indutry dummie Ye Ye Ye Ye Ye n F ** 5.41** 4.92** 3.11** 2.98** F ** 3.78** 2.36** 4.50** 4.92** R e Note: Heterocedaticity-conitent.e. in bracket. *: tatiticay ignificant at 10% eve; **: tatiticay ignificant at 5% eve. F 1 : joint ignificance of expanatory variabe. F 2 : joint ignificance of indutry dummie. 21

23 Tabe 4. Wage diperion. Dependent variabe: Wage diperion (1) (2) (3) SIZE ** ( ) UNION ** (5.035) TASK RATIO (39.217) MINICOMPUTERS 6.078* (3.693) ** ( ) ** (5.071) (22.545) 8.196** (3.422) ** ( ) ** (4.957) (35.029) TASK RATIO * MINICOMPUTERS (24.481) (24.694) NETWORK 8.556* (4.706) 9.864** (4.393) TASK RATIO * NETWORK (37.043) (36.343) Indutry dummie Ye Ye Ye n F ** 5.48** 5.66** F * * R e Note: Heterocedaticity-conitent.e. in bracket. *: tatiticay ignificant at 10% eve; **: tatiticay ignificant at 5% eve. F 1 : joint ignificance of expanatory variabe. F 2 : joint ignificance of indutry dummie. 22

24 Reference Abod, Jon M., Franci Kramarz and David N. Margoi (1999), "Hig Wage Worker and Hig Wage Firm", Econometrica, Vo. 67, No. 2, Marc, pp Daron Acemogu (1999), "Cange in Unempoyment and Wage Inequaity: An Aternative Teory and Some Evidence", American Economic Revie, vo.89, no.5, December, p Autor, David H., Larence F. Katz and Aan B. Krueger (1998), "Computing Inequaity: Have Computer Canged te Labor Market?", Quartery Journa of Economic, Vo. 113, Iue 4, November, pp Barte, Ann P., and Nacum Sicerman (1999), "Tecnoogica Cange and Wage: An Interindutry Anayi", Journa of Poitica Economy, Vo. 107, No. 2, Apri, pp Berman, Ei, Jon Bound and Zvi Griice (1994), "Cange in te Demand for Skied Labor Witin U.S. Manufacturing: Evidence from te Annua Survey of Manufacturer", Quartery Journa of Economic, Vo. 109, No. 2, May, pp Berman, Ei, Jon Bound and Stepen Macin (1998), "Impication of Ski-Biaed Tecnoogica Cange: Internationa Evidence", Quartery Journa of Economic, Voume 113, Iue 4, November, pp Brenaan, Timoty (1999), "Computeriation and age diperion: an anaytica reinterpretation", Economic Journa, Vo.109, No.456, June, pp Buo, Jeremy, and Larence H. Summer (1986), "A Teory of Dua Labor Market it Appication to Indutria Poicy, Dicrimination, and Keyneian Unempoyment", Journa of Labor Economic, 4, pp Damazzo, Aberto (2000), Tecnoogica Compexity, Wage Differentia and Unempoyment, mimeo, Department of Economic, Univerity of Siena. Davi, Steve J., and Jon Hatianger (1991), Wage Diperion beteen and itin U.S. Manufacturing Pant, , Brooking Paper on Economic Activity: Microeconomic, pp DiNardo, Jon E., and Jörn-Steffen Picke (1997), "Te Return to Computer Ue Reviited: Have Penci Canged te Wage Structure Too?", Quartery Journa of Economic, Vo. 112, Iue 1, February, pp Dom, Mark, Timoty Dunne, and Kennet R. Troke (1997), "Worker, Wage, and Tecnoogy", Quartery Journa of Economic, Vo.112, Iue 1, February, pp

25 Dunne, Timoty, Lucia Foter, Jon Hatianger, Kennet Troke (2000), Wage and Productivity Diperion in U.S. Manufacturing: Te Roe of Computer Invetment, NBER WP No Dunne, Timoty, and Jame A. Scmitz Jr. (1995), "Wage, Empoyment Structure and Empoyer Size-Wage Premia: Teir Reationip to Advanced-tecnoogy Uage at US Manufacturing Etabiment", Economica, Vo. 62, pp Entorf, Hort, Mice Goac, and Franci Kramarz (1999), "Ne Tecnoogie, Wage, and Worker Seection", Journa of Labor Economic, Vo. 17, No. 3, pp Gibbon, Robert S., and Larence F. Katz (1992), " Doe unmeaured abiity expain interindutry age differentia?", Revie of Economic Studie, Vo.59, No.3(200), Juy, pp Godin, Caudia, and Larence F. Katz (1996), "Tecnoogy, Ski, and te Wage Structure: Inigt from te Pat", American Economic Revie, Paper and Proceeding, Vo. 86, No. 2, May, pp Godin, Caudia, and Larence F. Katz (1998), "Te Origin of Tecnoogy-Ski Compementarity", Quartery Journa of Economic, Vo. 93, No. 3, Augut, pp Peter Gottcak (1997), "Inequaity, Income Grot, and Mobiity: Te Baic Fact", Journa of Economic Perpective, vo.11, no.2, Spring, p Hake, Jonatan, and Yva Heden (1999), "Computer and te Demand for Skied Labour: Indutry and Etabiment-Leve Pane Evidence for te UK", Economic Journa, Vo. 109, Marc, pp. C68-C79. Jonon, George E. (1997), Cange in Earning Inequaity: Te Roe of Demand Sift, Journa of Economic Perpective, Vo.11, No.2, Spring, pp Katz, Larence F. (1986), "Efficiency Wage Teorie: A Partia Evauation", NBER Macroeconomic Annua 1986, MIT Pre, Cambridge (MA). Kramarz, Franci, Stefan Loivier, and Loui-Pau Peè (1996), Wage Inequaitie and Firm- Specific Compenation Poicie in France, Annae d Economie et de Statitique, Vo , January-June, Kremer, Micae (1993), "Te O-Ring Teory of Economic Deveopment", Quartery Journa of Economic, Vo. 108, Iue 3, Augut, pp Kremer, Micae, and Erik Makin (1996), "Wage Inequaity and Segregation by Ski", NBER Working Paper No. 5718, Cambridge (MA). Revied verion: Marc

26 Krueger, Aan B. (1993), "Ho Computer Have Canged te Wage Structure: Evidence from Microdata, ", Quartery Journa of Economic, Vo. 108, Iue 1, February, pp Krueger, Aan B. and Larence H. Summer (1987), Refection on te Inter-Indutry Wage Structure. In Kevin Lang and Jonatan S. Leonard, ed., Unempoyment and te Structure of Labor Market, pp Bai Backe, Oxford. Krueger, Aan B., and Larence H. Summer (1988), " Efficiency age and te inter-indutry age tructure", Econometrica, Vo.56, No.2, Marc, pp Layard, Ricard, Stepen Nicke and Ricard Jackman (1991), Unempoyment. Macroeconomic Performance and te Labour Market, Oxford Univerity Pre, Oxford. Macin, Stepen, and Jon van Reenen (1998), "Tecnoogy and Cange in Ski Structure: Evidence from Seven OECD Countrie ", Quartery Journa of Economic, Vo. 113, Iue 4, November, pp Mark, Jerome A. (1987), Tecnoogica Cange and Empoyment: Some Reut from BLS Reearc, Monty Labor Revie, 110(4), pp Migrom, Pau, and Jon Robert (1990), "Te Economic of Modern Manufacturing: Tecnoogy, Strategy, and Organization", American Economic Revie, Vo. 80, No. 3, June, pp Migrom, Pau, and Jon Robert (1995), Compementaritie and Fit. Strategy, Structure, and Organizationa Cange in Manufacturing, Journa of Accounting and Economic, Vo. 19, pp Van Reenen, Jon (1996), "Te Creation and Capture of Rent: Wage and Innovation in a Pane of UK Companie", Quartery Journa of Economic, Vo. 111, Iue 1, February, pp Summer, Larence H. (1988), Reative Wage, Efficiency Wage, and Keyneian Unempoyment, American Economic Revie Paper and Proceeding, Vo. 78, No. 2, May, pp